As a method of applying optimal image processing according to features of an image, a method of extracting a character edge using a differential filter or the like to segment an image region into a text region and other regions, and then applying adaptive processing is known. Also, a method of extracting a halftone dot region by defining a dot as an isolation amount and integrating isolation amounts within a predetermined area, so as to segment an image region into a text region and the halftone dot region, and then applying adaptive processing is known (for example, see Japanese Patent Laid-Open No. 11-85978).
The aforementioned region extraction technique exhibits high extraction performance for a halftone dot region that expresses a character, photo, or the like in a white background. However, the extraction performance of a halftone dot region when a character, line drawing, or the like is present within the halftone dot region is not so high. Especially, it is very difficult to extract a low-density halftone dot region including a character or line drawing.
Extraction errors increase if processing is changed in a direction to make it easy to extract dots in order to improve the extraction performance of a halftone dot region. Also, a method of avoiding the influence of a character or line drawing by enlarging an area in which isolation amounts are integrated may be used. However, at the boundary of a halftone dot region that contacts a white background, the extraction result of the halftone dot region undesirably enlarges (the extraction result balloons).